Steps and workflow:

0 - Using this notebook

1 - Download data

2 - Explore data

3 - Choose and train a model

4 - Scoring

5 - Make prediction

import os
import numpy as np
import pandas as pd
import xgboost as xgb
import matplotlib.pyplot as plt
from xgboost import plot_importance, plot_tree
from sklearn.metrics import mean_squared_error
from sklearn.preprocessing import MinMaxScaler
from sklearn.model_selection import train_test_split, GridSearchCV

# Time series decomposition
#!pip install stldecompose
#from stldecompose import decompose
from statsmodels.tsa.seasonal import STL

# Chart drawing
import plotly as py
import plotly.io as pio
import plotly.graph_objects as go
from plotly.subplots import make_subplots
from plotly.offline import download_plotlyjs, init_notebook_mode, plot, iplot

# Mute sklearn warnings
from warnings import simplefilter
simplefilter(action='ignore', category=FutureWarning)
simplefilter(action='ignore', category=DeprecationWarning)

# Show charts when running kernel
init_notebook_mode(connected=True)

# Change default background color for all visualizations
layout=go.Layout(paper_bgcolor='rgba(0,0,0,0)', plot_bgcolor='rgba(250,250,250,0.8)')
fig = go.Figure(layout=layout)
templated_fig = pio.to_templated(fig)
pio.templates['my_template'] = templated_fig.layout.template
pio.templates.default = 'my_template'

1 - Download data

We will provide you with two dataset

Training_data will be use to train your model. Hackathon_data will be use to make your prediciton. There is three target you need to provide prediction on: target_r, target_g, target_b.

# Data Download (may take a few minutes depending on your network)
train_datalink_X = 'https://tournament.crunchdao.com/data/X_train.csv'  
train_datalink_y = 'https://tournament.crunchdao.com/data/y_train.csv' 
hackathon_data_link = 'https://tournament.crunchdao.com/data/X_test.csv' 

# Data for training
train_data = pd.read_csv(train_datalink_X)
# Data for which you will submit your prediction
test_data = pd.read_csv(hackathon_data_link)
# Targets use for your supervised trainning
train_targets = pd.read_csv(train_datalink_y)

# If you don't want to work with time series
train_data = train_data.drop(columns=['id'])
#test_data = test_data.drop(columns=['Moons', 'id'])

display(train_data)
display(train_targets)
display(test_data)

	Moons	Feature_1	Feature_2	Feature_3	Feature_4	Feature_5	Feature_6	Feature_7	Feature_8	Feature_9	...	Feature_19	Feature_20	Feature_21	Feature_22	Feature_23	Feature_24	Feature_25	Feature_26	Feature_27	Feature_28
0	0	0.00	0.25	0.25	0.00	0.25	0.25	0.25	0.25	0.25	...	0.50	0.50	0.25	1.00	0.25	0.00	0.25	0.50	0.25	0.75
1	0	1.00	0.50	0.75	0.75	0.75	0.75	0.75	0.75	1.00	...	0.50	0.50	0.25	0.50	0.25	0.00	0.25	1.00	0.25	0.50
2	0	0.25	1.00	0.00	0.00	0.75	0.75	0.25	0.25	0.75	...	0.50	0.75	0.00	0.25	0.25	0.75	1.00	0.25	1.00	0.75
3	0	0.25	0.00	0.50	0.50	1.00	1.00	0.50	0.50	0.25	...	0.50	0.00	0.25	0.50	0.25	0.50	1.00	1.00	0.25	0.75
4	0	1.00	0.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00	...	0.50	1.00	0.25	0.75	0.25	0.75	0.25	0.50	0.25	0.50
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
67892	28	1.00	1.00	0.00	0.50	0.75	0.75	0.50	0.50	0.75	...	0.25	0.00	0.25	1.00	1.00	1.00	0.75	0.50	0.25	0.50
67893	28	1.00	0.75	0.25	0.25	0.75	0.75	0.25	0.25	0.25	...	0.00	0.00	1.00	1.00	0.00	0.50	0.00	0.25	0.00	0.50
67894	28	1.00	1.00	1.00	0.25	0.00	0.00	0.75	0.75	1.00	...	0.25	1.00	1.00	0.25	0.00	0.50	0.25	0.25	1.00	0.25
67895	28	1.00	1.00	0.50	0.50	0.25	0.25	0.00	0.00	0.00	...	1.00	1.00	0.25	1.00	1.00	1.00	0.00	0.50	0.00	0.25
67896	28	0.50	0.50	0.25	0.25	0.75	0.75	0.75	0.75	0.75	...	0.25	1.00	1.00	0.00	1.00	0.50	1.00	1.00	0.00	0.00

67897 rows × 29 columns

	target_r	target_g	target_b
0	0.25	0.50	0.25
1	0.25	0.25	0.75
2	0.75	0.50	0.75
3	0.25	0.25	0.25
4	0.75	1.00	1.00
...	...	...	...
67892	0.50	0.50	0.50
67893	0.25	0.00	0.25
67894	0.50	0.25	0.25
67895	0.00	0.50	0.25
67896	0.00	0.00	0.25

67897 rows × 3 columns

	id	Moons	Feature_1	Feature_2	Feature_3	Feature_4	Feature_5	Feature_6	Feature_7	Feature_8	...	Feature_19	Feature_20	Feature_21	Feature_22	Feature_23	Feature_24	Feature_25	Feature_26	Feature_27	Feature_28
0	0021949c0f40bf7b8de636d5557a36bd	0	1.00	0.75	0.75	0.75	0.50	0.0	0.75	0.0	...	1.00	1.00	0.00	0.75	0.25	0.75	0.25	0.00	1.00	0.75
1	0044a1eb91e6d9e766bfc070a58c091e	0	0.00	0.75	1.00	1.00	0.50	0.0	0.50	0.0	...	0.50	1.00	0.25	0.75	0.25	0.75	1.00	0.25	0.00	0.25
2	004b477e43f22d3eb0b7e318badb2438	0	0.00	0.00	0.75	0.75	0.25	0.0	0.25	0.0	...	0.50	0.50	0.25	0.00	0.25	0.00	0.25	1.00	0.25	1.00
3	005e227ee57221853cc374b5d5d79256	0	0.00	0.50	0.75	0.75	0.50	0.0	0.75	0.0	...	0.00	0.00	0.25	0.75	0.00	0.75	1.00	1.00	1.00	1.00
4	006b9832e7683a7ec2b6c3d177c4715a	0	0.00	1.00	0.75	0.75	0.25	0.0	0.25	0.0	...	0.25	1.00	0.25	0.75	0.00	0.75	0.25	0.25	0.75	0.25
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2223	ff9baadf910a626a35599a1c15ddb42a	0	0.75	1.00	0.50	0.50	0.25	1.0	1.00	1.0	...	0.00	0.50	0.00	0.50	0.00	0.25	1.00	0.50	0.75	0.75
2224	ffafd9c92b60f98962902f7cae7e3b66	0	0.50	0.25	0.00	0.00	0.75	1.0	0.75	1.0	...	1.00	0.25	1.00	0.25	1.00	0.00	0.00	1.00	0.75	1.00
2225	ffc01cf716931e050512a9a106150a40	0	1.00	0.50	0.00	0.00	0.75	1.0	0.00	1.0	...	0.00	0.75	1.00	0.00	0.00	0.00	0.75	0.25	0.75	0.25
2226	ffe524c3576df2a4b9cb3acd13fbb355	0	1.00	0.25	0.25	0.25	0.50	1.0	0.50	1.0	...	1.00	0.50	1.00	0.25	1.00	0.00	1.00	0.75	0.75	0.25
2227	ffe8a6d4e8f7efc676f60b8ae9d8481d	0	0.50	0.00	0.75	0.25	0.25	1.0	0.25	1.0	...	1.00	0.50	1.00	1.00	1.00	1.00	0.75	0.50	0.75	0.00

2228 rows × 30 columns

display(train_data.describe())
display(train_targets.describe())

	Moons	Feature_1	Feature_2	Feature_3	Feature_4	Feature_5	Feature_6	Feature_7	Feature_8	Feature_9	...	Feature_19	Feature_20	Feature_21	Feature_22	Feature_23	Feature_24	Feature_25	Feature_26	Feature_27	Feature_28
count	67897.00000	67897.000000	67897.000000	67897.000000	67897.000000	67897.000000	67897.000000	67897.000000	67897.000000	67897.000000	...	67897.000000	67897.000000	67897.000000	67897.000000	67897.000000	67897.000000	67897.000000	67897.000000	67897.000000	67897.000000
mean	13.79625	0.499985	0.499985	0.499985	0.499985	0.499985	0.499985	0.499985	0.499985	0.499985	...	0.499985	0.499985	0.499985	0.499985	0.499985	0.499985	0.499985	0.499985	0.499985	0.499985
std	8.46247	0.353621	0.353621	0.353621	0.353621	0.353621	0.353621	0.353621	0.353621	0.353621	...	0.353621	0.353621	0.353621	0.353621	0.353621	0.353621	0.353621	0.353621	0.353621	0.353621
min	0.00000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
25%	6.00000	0.250000	0.250000	0.250000	0.250000	0.250000	0.250000	0.250000	0.250000	0.250000	...	0.250000	0.250000	0.250000	0.250000	0.250000	0.250000	0.250000	0.250000	0.250000	0.250000
50%	14.00000	0.500000	0.500000	0.500000	0.500000	0.500000	0.500000	0.500000	0.500000	0.500000	...	0.500000	0.500000	0.500000	0.500000	0.500000	0.500000	0.500000	0.500000	0.500000	0.500000
75%	21.00000	0.750000	0.750000	0.750000	0.750000	0.750000	0.750000	0.750000	0.750000	0.750000	...	0.750000	0.750000	0.750000	0.750000	0.750000	0.750000	0.750000	0.750000	0.750000	0.750000
max	28.00000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	...	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000

8 rows × 29 columns

	target_r	target_g	target_b
count	67897.000000	67897.000000	67897.000000
mean	0.499985	0.499985	0.499985
std	0.353621	0.353621	0.353621
min	0.000000	0.000000	0.000000
25%	0.250000	0.250000	0.250000
50%	0.500000	0.500000	0.500000
75%	0.750000	0.750000	0.750000
max	1.000000	1.000000	1.000000

display(train_data.info())

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 67897 entries, 0 to 67896
Data columns (total 29 columns):
 #   Column      Non-Null Count  Dtype  
---  ------      --------------  -----  
 0   Moons       67897 non-null  int64  
 1   Feature_1   67897 non-null  float64
 2   Feature_2   67897 non-null  float64
 3   Feature_3   67897 non-null  float64
 4   Feature_4   67897 non-null  float64
 5   Feature_5   67897 non-null  float64
 6   Feature_6   67897 non-null  float64
 7   Feature_7   67897 non-null  float64
 8   Feature_8   67897 non-null  float64
 9   Feature_9   67897 non-null  float64
 10  Feature_10  67897 non-null  float64
 11  Feature_11  67897 non-null  float64
 12  Feature_12  67897 non-null  float64
 13  Feature_13  67897 non-null  float64
 14  Feature_14  67897 non-null  float64
 15  Feature_15  67897 non-null  float64
 16  Feature_16  67897 non-null  float64
 17  Feature_17  67897 non-null  float64
 18  Feature_18  67897 non-null  float64
 19  Feature_19  67897 non-null  float64
 20  Feature_20  67897 non-null  float64
 21  Feature_21  67897 non-null  float64
 22  Feature_22  67897 non-null  float64
 23  Feature_23  67897 non-null  float64
 24  Feature_24  67897 non-null  float64
 25  Feature_25  67897 non-null  float64
 26  Feature_26  67897 non-null  float64
 27  Feature_27  67897 non-null  float64
 28  Feature_28  67897 non-null  float64
dtypes: float64(28), int64(1)
memory usage: 15.0 MB

None

Moving Averages

train_data['EMA_9'] = train_data['Feature_2'].ewm(9).mean().shift()
train_data['EMA_5'] = train_data['Feature_2'].rolling(5).mean().shift()
train_data['EMA_10'] = train_data['Feature_2'].rolling(10).mean().shift()
train_data['EMA_15'] = train_data['Feature_2'].rolling(15).mean().shift()
train_data['EMA_30'] = train_data['Feature_2'].rolling(30).mean().shift()

fig = go.Figure()
fig.add_trace(go.Scatter(x=train_data.Moons, y=train_data.EMA_9, name='EMA 9'))
fig.add_trace(go.Scatter(x=train_data.Moons, y=train_data.EMA_5, name='EMA 5'))
fig.add_trace(go.Scatter(x=train_data.Moons, y=train_data.EMA_10, name='EMA 10'))
fig.add_trace(go.Scatter(x=train_data.Moons, y=train_data.EMA_15, name='EMA 15'))
fig.add_trace(go.Scatter(x=train_data.Moons, y=train_data.EMA_30, name='EMA 30'))
fig.show()

RSI

def relative_strength_idx(df, n=14):
    close = train_data['Feature_2']
    delta = close.diff()
    delta = delta[1:]
    pricesUp = delta.copy()
    pricesDown = delta.copy()
    pricesUp[pricesUp < 0] = 0
    pricesDown[pricesDown > 0] = 0
    rollUp = pricesUp.rolling(n).mean()
    rollDown = pricesDown.abs().rolling(n).mean()
    rs = rollUp / rollDown
    rsi = 100.0 - (100.0 / (1.0 + rs))
    return rsi

train_data['RSI'] = relative_strength_idx(train_data).fillna(0)

fig = go.Figure(go.Scatter(x=train_data.Moons, y=train_data.RSI, name='RSI'))
fig.show()

MACD

EMA_12 = pd.Series(train_data['Feature_2'].ewm(span=12, min_periods=12).mean())
EMA_26 = pd.Series(train_data['Feature_2'].ewm(span=26, min_periods=26).mean())
train_data['MACD'] = pd.Series(EMA_12 - EMA_26)
train_data['MACD_signal'] = pd.Series(train_data.MACD.ewm(span=9, min_periods=9).mean())

fig = make_subplots(rows=2, cols=1)
fig.add_trace(go.Scatter(x=train_data.Moons, y=train_data.Feature_1, name='Feature_2'), row=1, col=1)
fig.add_trace(go.Scatter(x=train_data.Moons, y=EMA_12, name='EMA 12'), row=1, col=1)
fig.add_trace(go.Scatter(x=train_data.Moons, y=EMA_26, name='EMA 26'), row=1, col=1)
fig.add_trace(go.Scatter(x=train_data.Moons, y=train_data['MACD'], name='MACD'), row=2, col=1)
fig.add_trace(go.Scatter(x=train_data.Moons, y=train_data['MACD_signal'], name='Signal line'), row=2, col=1)
fig.show()

Split data frame into three subsets: trainin(70%), validation(15%), test(15%)

test_size = 0.15
valid_size = 0.15

test_split_idx  = int(train_data.shape[0] * (1-test_size))
valid_split_idx = int(train_data.shape[0] * (1-(valid_size+test_size)))

train_df  = train_data.loc[:valid_split_idx].copy()
valid_df  = train_data.loc[valid_split_idx+1:test_split_idx].copy()
test_df   = train_data.loc[test_split_idx+1:].copy()

fig = go.Figure()
fig.add_trace(go.Scatter(x=train_df.Moons, y=train_df.Feature_2, name='Training'))
fig.add_trace(go.Scatter(x=valid_df.Moons, y=valid_df.Feature_2, name='Validation'))
fig.add_trace(go.Scatter(x=test_df.Moons,  y=test_df.Feature_2,  name='Test'))
fig.show()

drop columns not being used

drop_cols = ['Moons', 'Feature_1', 'Feature_3', 'Feature_4', 'Feature_5', 'Feature_6', 'Feature_7', 'Feature_8', 'Feature_9', 'Feature_10', 'Feature_11', 'Feature_12', 'Feature_13', 'Feature_14', 'Feature_15', 'Feature_16', 'Feature_17', 'Feature_18', 'Feature_19', 'Feature_20', 'Feature_21', 'Feature_22', 'Feature_23', 'Feature_24', 'Feature_25', 'Feature_26', 'Feature_27', 'Feature_28']

train_df = train_df.drop(drop_cols, 1)
valid_df = valid_df.drop(drop_cols, 1)
test_df  = test_df.drop(drop_cols, 1)

Split into features and labels

y_train = train_df['Feature_2'].copy()
X_train = train_df.drop(['Feature_2'], 1)

y_valid = valid_df['Feature_2'].copy()
X_valid = valid_df.drop(['Feature_2'], 1)

y_test  = test_df['Feature_2'].copy()
X_test  = test_df.drop(['Feature_2'], 1)

X_train.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 47528 entries, 0 to 47527
Data columns (total 8 columns):
 #   Column       Non-Null Count  Dtype  
---  ------       --------------  -----  
 0   EMA_9        47527 non-null  float64
 1   EMA_5        47523 non-null  float64
 2   EMA_10       47518 non-null  float64
 3   EMA_15       47513 non-null  float64
 4   EMA_30       47498 non-null  float64
 5   RSI          47527 non-null  float64
 6   MACD         47503 non-null  float64
 7   MACD_signal  47495 non-null  float64
dtypes: float64(8)
memory usage: 2.9 MB

%%time

parameters = {
    'n_estimators': [400],
    'learning_rate': [0.05],
    'max_depth': [15],
    'gamma': [0.02],
    'random_state': [42]
}

eval_set = [(X_train, y_train), (X_valid, y_valid)]
model = xgb.XGBRegressor(eval_set=eval_set, objective='reg:squarederror')
clf = GridSearchCV(model, parameters)

clf.fit(X_train, y_train)

print(f'Best params: {clf.best_params_}')
print(f'Best validation score = {clf.best_score_}')

[15:38:44] WARNING: ..\src\learner.cc:541: 
Parameters: { eval_set } might not be used.

  This may not be accurate due to some parameters are only used in language bindings but
  passed down to XGBoost core.  Or some parameters are not used but slip through this
  verification. Please open an issue if you find above cases.


[15:39:25] WARNING: ..\src\learner.cc:541: 
Parameters: { eval_set } might not be used.

  This may not be accurate due to some parameters are only used in language bindings but
  passed down to XGBoost core.  Or some parameters are not used but slip through this
  verification. Please open an issue if you find above cases.


[15:40:27] WARNING: ..\src\learner.cc:541: 
Parameters: { eval_set } might not be used.

  This may not be accurate due to some parameters are only used in language bindings but
  passed down to XGBoost core.  Or some parameters are not used but slip through this
  verification. Please open an issue if you find above cases.


[15:41:28] WARNING: ..\src\learner.cc:541: 
Parameters: { eval_set } might not be used.

  This may not be accurate due to some parameters are only used in language bindings but
  passed down to XGBoost core.  Or some parameters are not used but slip through this
  verification. Please open an issue if you find above cases.


[15:42:31] WARNING: ..\src\learner.cc:541: 
Parameters: { eval_set } might not be used.

  This may not be accurate due to some parameters are only used in language bindings but
  passed down to XGBoost core.  Or some parameters are not used but slip through this
  verification. Please open an issue if you find above cases.


[15:43:36] WARNING: ..\src\learner.cc:541: 
Parameters: { eval_set } might not be used.

  This may not be accurate due to some parameters are only used in language bindings but
  passed down to XGBoost core.  Or some parameters are not used but slip through this
  verification. Please open an issue if you find above cases.


Best params: {'gamma': 0.02, 'learning_rate': 0.05, 'max_depth': 15, 'n_estimators': 400, 'random_state': 42}
Best validation score = 0.8524452403640194
Wall time: 6min 12s

%%time

model = xgb.XGBRegressor(**clf.best_params_, objective='reg:squarederror')
model.fit(X_train, y_train, eval_set=eval_set, verbose=False)

Wall time: 1min 16s

XGBRegressor(base_score=0.5, booster='gbtree', colsample_bylevel=1,
             colsample_bynode=1, colsample_bytree=1, gamma=0.02, gpu_id=-1,
             importance_type='gain', interaction_constraints='',
             learning_rate=0.05, max_delta_step=0, max_depth=15,
             min_child_weight=1, missing=nan, monotone_constraints='()',
             n_estimators=400, n_jobs=12, num_parallel_tree=1, random_state=42,
             reg_alpha=0, reg_lambda=1, scale_pos_weight=1, subsample=1,
             tree_method='exact', validate_parameters=1, verbosity=None)

plot_importance(model);

Calculate and predict prices

y_pred = model.predict(X_test)
print(f'y_true = {np.array(y_test)[:5]}')
print(f'y_pred = {y_pred[:5]}')

y_true = [0.25 0.5  0.   0.   0.  ]
y_pred = [ 0.13695318  0.3457684   0.06881184  0.0295731  -0.01050354]

print(f'mean_squared_error = {mean_squared_error(y_test, y_pred)}')

mean_squared_error = 0.018134200519696882

predicted_prices = train_data.loc[test_split_idx+1:].copy()
predicted_prices['Feature_1'] = y_pred

fig = make_subplots(rows=2, cols=1)
fig.add_trace(go.Scatter(x=train_data.Moons, y=train_data.Feature_1,
                         name='Truth',
                         marker_color='LightSkyBlue'), row=1, col=1)

fig.add_trace(go.Scatter(x=predicted_prices.Moons,
                         y=predicted_prices.Feature_1,
                         name='Prediction',
                         marker_color='MediumPurple'), row=1, col=1)

fig.add_trace(go.Scatter(x=predicted_prices.Moons,
                         y=y_test,
                         name='Truth',
                         marker_color='LightSkyBlue',
                         showlegend=False), row=2, col=1)

fig.add_trace(go.Scatter(x=predicted_prices.Moons,
                         y=y_pred,
                         name='Prediction',
                         marker_color='MediumPurple',
                         showlegend=False), row=2, col=1)

fig.show()